create_refund
AI agents use create_refund to commit financial operations through AdminAgent — usually the final step of a payment, billing, or trading workflow. A call moves real money.
Refunds are financial transactions that reduce revenue and move money out of the business to customers. This is the most severe category per the rule hierarchy (Financial > Destructive > Execute > Write > Read). The blast radius is critical: an agent creating refunds without proper authorization or for fraudulent purposes could cause significant financial loss.
From the tool's definition Tool name is 'create_refund' which directly involves moving money back to customers. Shopify refunds reverse financial transactions and return funds to payment methods. Context is Shopify Admin with 'full store management' capabilities.
Attacks that exploit this kind of access
create_refund. It is categorised as a Financial tool in the AdminAgent MCP Server, which means it involves financial transactions. Block by default and require explicit approval.
Register the AdminAgent MCP server in PolicyLayer and add a rule for create_refund: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches AdminAgent. Nothing to install.
create_refund is a Financial tool with critical risk. Critical-risk tools should be blocked by default and only enabled with explicit human approval.
Yes. Add a rate_limit block to the create_refund rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for create_refund. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
create_refund is provided by the AdminAgent MCP server (rushikeshmore/admin-agent). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →